Video optimization can be utilized in a service provider network to improve consumer viewing experience by, for example, reducing start times, reducing frequency of re-buffering events, or the like. Video optimization can also aim to reduce an amount of network bandwidth consumed by a video traffic flow. Thus, in order to perform video optimization, the service provider needs to identify video traffic flows in the service provider network. Typical techniques for identifying video traffic flows include, for example, deep packet inspection, using a model trained on a dataset, and/or the like. However, such techniques can be inaccurate in practice.